I have an XML file all in one stream, without any Carriage Return / Line Feeds. The file looks something like this:

<tag1>data</tag1><tag2>data2</tag2><tag3>etc</tag3>

Where there is a ><, I need to insert a new line in between the >< so that the file looks like:

<tag1>data</tag1>
<tag2>data2</tag2>
<tag3>etc</tag3>

myfile contains <tag1>data</tag1><tag2>data2</tag2><tag3>etc</tag3>

filename=$myfile
cat $filename | sed 's|><|>\<|'g > new

new was the same as $filename

Then I treid cat $filename | sed 's|><|\<|'g > new

cat new responded with
<tag1>data</tag1<tag2>data2</tag2<tag3>etc</tag3<

What am I doing wrong?

Thanks!!

Not thoroughly tested! But maybe try:

That will edit the file in-place, so make a backup of it before proceeding. It looks for the closing tag, and replaces it with itself PLUS a newline. It works for me in the console, with a test file containing your example text, but in the real situation, it may not work, depending on what actual DATA is in your file between the tags.


I'm not sure fully what you're doing wrong in your first attempt(s) there -- you'd need to explain what your code was supposed to do, as I don't quite understand it.

How about:

When I do the command below on the command line

cat $filename | sed -i 's%\(</tag[0-9]>\)%\1\n%g' > testfile

I get the following error:
sed: illegal option -- i



When I do
cat $filename | sed 's/></>\n</g' > testfile

I get n between the ><

What is your sed version

@ linuxready -- you are not supposed to CAT the file into the command I gave; the "-i" option specifies an in-place operation, meaning that SED itself will read the file and operate on it; no cat required.

If you insist on using cat, then remove the -i from the sed statement, cat your file into the sed statement, and use a shell redirect to dump the output into a new file.

PS - NOTE, that "testfile" in my example, refers to the file you are operating on (I should have used the name "filename" instead -- sorry.)

Sorry!!! I just realized that I was working on the SunOS 5.10 Unix server and not our Linux server.

So sed --v, sed -V, or sed -version is not listing the version of sed on the Unix server.

Also,
I tried cat $filename | sed 's%\(</tag[0-9]>\)%\1\n%g' > new

and it put n in between >< for instance >n<

I've asked for this thread to be moved to the Solaris forum, where the help offered will hopefully be better suited to the OS you're dealing with.

Meanwhile, I suppose you need to examine whatever the differences are between GNU sed, and whatever sed you have at your disposal, and adjust the suggestion(s) given to see if you can produce the desired functionality. OR, (and this may be more trouble than it's worth) seek a method other than sed, for this task (though it is IMO definitely a sed job).

Moved: This thread is more suitable in <*Solaris> as that system uses antiques that don't have half the functionality of the GNU variants, and has been moved accordingly to help your thread/question get the exposure it deserves.

Found out that you have to place the newline literary.You have to press enter. Can't test it myself.
hm works also in a bash script.

@ whizje -- nice find! How primitive is that that a \n doesn't work..

Now we just need the OP to test for us, and see if it works in their environment.

Thanks,

Sasha

Hey, two LQ mods bashing Solaris in the same thread ...

Let me answer to that:

Solaris vs Gnu sed behavior differences are mainly due to the fact Solaris and Gnu developers have quite different rules, goals and priorities. The former are very serious about upward compatibility so are very conservative regarding changes that either might break existing scripts or might introduce standard compliance conflicts in the future. Sed is one of these commands having their interfaces classified as "Standard", i.e. committed to a standard defined outside Sun^H^H^HOracle. That mostly means if the standard doesn't evolve, there is little hope to see a request for enhancement implemented.

The "-i" option is missing from Solaris sed because it isn't defined by POSIX. Implementing it is too risky as it would create a conflict in the (quite unlikely) event that the standard decide one day that "-i" would serve a different purpose.

The way to split a line is specified by the POSIX standard that way:
Solaris sed is then supporting that way to do it. Implementing the "\n" extension would break existing scripts that expect that sequence to simply mean "n". It isn't going to happen.

In any case, there are several workarounds should you object to use raw newlines in the command.

The most common is simply to use Gnu sed instead of the default one. Gnu tools have been easily available for SunOS/Solaris for ages, even before Linux was born. Gnu sed is actually now part of Solaris distributions based on the OpenSolaris code base, usually under /usr/gnu/bin/sed.

Alternatively, you can use a different tool that is more friendly with "\n" like awk, perl or whatever, eg:

Two good places for sed information are the unix grymoire and the sed faq. Both cover multiple versions of the program.

http://www.grymoire.com/Unix/Sed.html
http://sed.sourceforge.net/sedfaq.html

The sed faq even has a section devoted specifically to embedding newlines:

http://sed.sourceforge.net/sedfaq4.html#s4.1